Real-time server visualization with canvas and processing.js(cloudkick.com)
cloudkick.com
Real-time server visualization with canvas and processing.js
https://www.cloudkick.com/blog/2010/apr/27/cloudkick_server_visualization_in_html5_canvas/
8 comments
Disclaimer: I work at Cloudkick.
Imagine looking at a list of 500 servers and trying to determine how their aggregate memory usage was trending in real-time.
To quickly address your concerns:
1. Servers that are critical turn red and blink - there's no need to filter anything since your eyes have already done it.
2. This tool isn't really designed to be a gateway to arbitrary servers; it's showing you anomalies, problems, and aggregate trending.
3. I think this could be a fair point, but with only 3 (clearly labeled) axes, I think it's hard to suggest that the learning curve for digestion the information is incredibly steep. If you actually think how it's set up at the moment is too confusing, I'd love to hear suggestions.
We leave this up on the big screen at the office and it makes it very easy to tell when something untoward is happening with our servers, without really having to devote direct attention to the visualization. We designed it so problems stand out and make themselves known to you.
One of our guiding principles is to move quickly and take user feedback though, so if you've got any constructive criticism, let's keep the discussion going!!!
Imagine looking at a list of 500 servers and trying to determine how their aggregate memory usage was trending in real-time.
To quickly address your concerns:
1. Servers that are critical turn red and blink - there's no need to filter anything since your eyes have already done it.
2. This tool isn't really designed to be a gateway to arbitrary servers; it's showing you anomalies, problems, and aggregate trending.
3. I think this could be a fair point, but with only 3 (clearly labeled) axes, I think it's hard to suggest that the learning curve for digestion the information is incredibly steep. If you actually think how it's set up at the moment is too confusing, I'd love to hear suggestions.
We leave this up on the big screen at the office and it makes it very easy to tell when something untoward is happening with our servers, without really having to devote direct attention to the visualization. We designed it so problems stand out and make themselves known to you.
One of our guiding principles is to move quickly and take user feedback though, so if you've got any constructive criticism, let's keep the discussion going!!!
We do pretty exactly what you described on the monitoring overview -- except for the filter in #1. Great feedback, however, and we will look at including that!
This is a very different take on visualizing your infrastructure. I encourage you to setup your servers and post on the big screen at your office. After watching it for awhile, you will start to be able to see when something is an anomaly just with a glance.
This is a very different take on visualizing your infrastructure. I encourage you to setup your servers and post on the big screen at your office. After watching it for awhile, you will start to be able to see when something is an anomaly just with a glance.
there is a gridview inside cloudkick, this is just an alternative view on the data, for example this is the dashboard:
https://www.cloudkick.com/site_media/images/home_tour/dashbo...
There is also the /monitoring/ page (which i don't have a screenshot of handy), which is a grid view similar to what you are talking about.
There is also the /monitoring/ page (which i don't have a screenshot of handy), which is a grid view similar to what you are talking about.
I'm looking at the Mozilla example on my laptop without a mouse, and I get a little bit annoyed when I'm hovering a certain node trying to click on it and the visualization keeps on moving and makes it hard to click on it. So maybe it could be possible to have a simple option to freeze the view?
And maybe an easy way of instantly changing to some near optimal viewpoint for viewing the data in terms of a given axis? (Ie. CPU or memory usage).
But this, combined with normal alerts would be cool. Oh, and what about showing trace lines once I've selected a given node? :)
And maybe an easy way of instantly changing to some near optimal viewpoint for viewing the data in terms of a given axis? (Ie. CPU or memory usage).
But this, combined with normal alerts would be cool. Oh, and what about showing trace lines once I've selected a given node? :)
Neat tech, terrible interface - utterly pointless. Flat graphs, which are configured to show outliers/only problematic nodes would provide a far more efficient interface to someone trying to work out where the bottlenecks and issues are. Showing lots of data for the sake of it might be pretty but it's not helpful, choosing what to present is an important part of the visualisation development process as how it is presented.
It looks like it was fun to make, but I can't see how it's particularly useful. I can't see anything useful in a 3D visualization like that.
> "and it's the first cloud monitoring system in the world that you can watch in real-time as it checks your servers."
Yeah Cacti does that with an HTTP refresh every few seconds. 'realtime' enough for most instances.
> "and it's the first cloud monitoring system in the world that you can watch in real-time as it checks your servers."
Yeah Cacti does that with an HTTP refresh every few seconds. 'realtime' enough for most instances.
I saw this a while back at the sfcloudhackers hack day; simply amazing.
Consider :-
1. A list can quickly be filtered to show only critical state servers - I probably don't care about servers which aren't doing anything
2. If I want to look for a specific server - I have no idea which one it is - I have to click to find it. A grid means I can look for it instantly
3. The info being displayed takes a while to digest as to what I'm seeing on the 3d axes with regards to ping etc... A grid just gives me the numbers and I can read them quickly.
Yet again a use of pretty graphics where it really isn't needed